The method
How people learn to have AI do their work
The Multiply Academy programme teaches people to have AI do their work one piece at a time, and they practise on their own jobs. It has seven parts, called modules.
What people learn
People practise three skills on every task.
- Giving context. Context is what AI needs to know to do a job properly. That means who the client is, what has happened on the project so far, what the source material says and what good looks like in your business. Experienced people carry most of this in their heads. They learn to get it in front of their AI, and how much to give. If they give it everything at once the work gets worse, so they choose what each job needs.
- Giving direction. They learn to describe the result they want and leave the method to the AI. They point it at the right background and let it ask questions before it starts.
- Judging the result. They read what comes back, correct it and decide whether it is good enough. That checking takes time and attention, and nothing goes out until they have done it.
People who direct AI well are clear about what they want and why, and know good work when they see it.
The shared brain
As they work, each person builds what we call a shared brain. It is a set of notes that both they and their AI read and add to.
Your shared brain
- Module 1
- The people you work with
- Your current projects
- Your recent emails
- Examples of how you like things done
Module 2 adds no notes.
- Module 3
- A short introduction your AI reads first
- A note on each client
- Your products and processes
- Your priorities
- Module 4
- Notes you speak into your phone
- Notes from each meeting
- Saved instructions for jobs you repeat
- Module 5
- Your specialist knowledge, written down
- Module 6
- Plans for long projects
- Research you have gathered
- Module 7
- A library of the best sources in your field
The notes are plain files in a folder on their own computer. They hold what an experienced person knows about their clients, projects and standards, written down task by task. Their AI reads the relevant ones before it starts a job, so less has to be explained again.
Each task adds to the notes. When the person corrects their AI, the correction goes into the notes too, so the AI is less likely to make the same mistake next time. People check what gets saved, because the AI can save something wrong.
The files are plain text, so other AI tools can read them too, though tools differ in how well they use them. We teach the programme on Claude Code, an AI tool from Anthropic that works with the files on your computer.
Each person builds their own shared brain first. When their team is ready, we help it set up sharing in a folder the organisation already uses. This step is not one of the seven modules. Each person leaves out anything private. Then each person’s AI can draw on what the team has shared.
What happens over the seven modules
We see people go through three stages.
First they learn to manage context. In Modules 1 to 3 they set up the tools, build their shared brain and get their first real work done by AI.
Then working through AI becomes a habit. From Module 4, everyday work such as meeting notes, follow-ups and email runs through their AI, and they stop having to remember to use it.
Then some of them take on work they had never tried. Ben Hollebon, Director of Digital and Transformation at the Methodist Church in Britain, built working prototypes of an app and a CRM with no software engineering background. He says, “Not just about how much you can do, but now what’s possible.” Read Ben Hollebon’s story
How the learning is built
A lot of training stops at what AI can do, and people go back to working as before. So Multiply Academy built its programme on research into how people learn and what makes a new habit last. We call the result the Multiply Method, and it has ten principles.
1. Each reflection shapes the next module
At the end of each module the person writes a reflection. It starts with a few questions answered from memory. The platform checks the reflection against the module that person was given. The next module then opens with what they did well, and goes over the one or two things they missed before teaching anything new. There are no scores and no pass marks.
- A module written for you
- Your real work
- Your reflection
- What you missed, covered first
These four steps start again with each new module.
This draws on the education researchers Paul Black and Dylan Wiliam. They found that pupils in school learn more when the teacher checks what each one knows and teaches to that. It also draws on research by the psychologist Robert Bjork and others. They found that recalling something helps you remember it better than reading it again.
2. The exercises are the person’s real work
Training often fails at the point where people go back to their job and can’t apply what they learned. More than a century ago the psychologist Edward Thorndike found that a skill carries over best when the practice task and the real task have the same elements. In this programme the exercise is the real task. The first one names the client and the project the person told us about a few minutes earlier.
3. Each module is written for one person
The educational psychologist Benjamin Bloom studied school pupils who were tutored one to one and helped to master each step before the next. They did far better than pupils taught in a class. He also pointed out that nobody could afford to teach everyone that way. Each of our modules is generated for one person, from what they have told us about their job and how they got on with the module before. Everyone learns the same things, from examples taken from their own work. Each module opens when the one before is finished and reflected on.
Louis von Mandach owns elfero AG, a building-automation firm of about twelve people in Switzerland. He describes the programme as “like a coach that you have at your side”. Read Louis von Mandach’s story

4. The hard weeks are planned for
When people start using a new skill on real work, their confidence and their results often drop before they get better. The education researcher Michael Fullan studied this in schools and calls it the implementation dip. Many people take it as a sign they are failing and go back to the old way. Every module has a section on what the person is likely to be feeling at that point and what to do about it.
The psychologist Albert Bandura found that confidence comes mostly from succeeding at real tasks yourself, and next from seeing people like you succeed. So people’s first successes come on their own work, and teams can see what their colleagues are building.
The other six
- Early modules give complete worked examples, and from Module 3 the examples leave more for the person to do. This draws on the educational psychologist John Sweller’s work on how much a learner can hold in mind at once.
- The same four ideas come back in every module, in more depth each time. People give AI the context, keep it in the shared brain, let AI do the work and check the result. The cognitive psychologist Jerome Bruner called this a spiral.
- Help is withdrawn on a schedule. The first modules give exact steps and the last ones give only the goal. This follows the developmental psychologist Lev Vygotsky, and the education researcher Slava Kalyuga’s finding that help which suits a beginner gets in an expert’s way.
- Each module links the new material to what the person already knows, as the educational psychologist David Ausubel recommended.
- People experience the change before it is explained to them. In Module 1 the first result comes before the theory. This draws on the adult-learning theorist Jack Mezirow.
- Each module takes up a problem people tend to hit after the one before, such as notes that get messy. The learning scientist Manu Kapur found that people learn more from an explanation when they have first struggled with the problem.
This research is about how people learn in general. Much of it was done with school pupils and students. Nobody has yet studied our programme independently. The full paper names every researcher: The Multiply Method (PDF).
What we leave out
We don’t think points, speed drills or multiple-choice tests change how adults work, so the programme has none. It doesn’t tailor to learning styles either, because research finds they don’t predict how well people learn.
What it asks of each person
It takes about 25 to 35 hours over about ten weeks, all spent on real work. Each person needs to want to practise, to set time aside and to have their organisation’s support.
Louis von Mandach says there were times, while he was teaching his AI how he works, when “it takes more time to teach than what you get out of it.” Now he says, “I don’t know how I could do without it.”